Artificial Intelligence in Chest Computed Tomography Optimization: A Task-based Framework for Balancing Detectability and Radiation Dose.
Authors
Affiliations (5)
Affiliations (5)
- Department of Radiology and Imaging Technology, School of Allied Health Sciences, Datta Meghe Institute of Higher Education and Research, Wardha, Maharashtra, India.
- Department of Medical Physics and Radiation Safety, Datta Meghe Institute of Higher Education and Research, Wardha, Maharashtra, India.
- Department of Radiodiagnosis, Datta Meghe Institute of Higher Education and Research, Wardha, Maharashtra, India.
- Department of Cardiovascular and Thoracic Surgery, School of Allied Health Sciences, Datta Meghe Institute of Higher Education and Research, Wardha, Maharashtra, India.
- Department of Medical Radiology Imaging and Therapeutic Technology, Parul Institute of Allied and Healthcare Sciences, Parul University, Vadodara, Gujarat, India.
Abstract
Computed tomography (CT) of the chest is essential for thoracic imaging because of its high spatial resolution and broad clinical applications. However, variability in acquisition parameters and reconstruction techniques contributes to inconsistent radiation dose and image quality, while conventional optimization approaches remain limited by complex parameter interactions and operator dependence. Optimization in chest CT is increasingly recognized as a task-based challenge in which diagnostic performance depends on lesion detectability rather than global image quality metrics alone. This review synthesizes artificial intelligence (AI)-assisted approaches for optimizing chest CT protocols within a task-based framework, with emphasis on lesion detectability, diagnostic efficacy, and radiation dose reduction, while examining how AI can address protocol variability and improve patient safety. A structured narrative review of PubMed-indexed literature published between January 2015 and January 2025 was conducted, focusing on studies evaluating AI-assisted acquisition, deep learning reconstruction, and radiation dose optimization in adult chest CT. The review highlights that the most effective outcomes are achieved when AI-based reconstruction is integrated with dose optimization strategies such as automatic exposure control and patient-adapted acquisition, rather than applied in isolation. Although the magnitude of benefit varies according to diagnostic task, study design, and evaluation metrics, AI consistently reduces operator variability and promotes protocol standardization. Overall, AI enables task-based optimization of chest CT by prioritizing lesion detectability and diagnostic performance over traditional image quality metrics alone. By integrating acquisition, reconstruction, and continuous feedback mechanisms, AI has the potential to support consistent, patient-specific optimization while maintaining diagnostic efficacy when appropriately validated.